Every AI leaderboard

Leaderboards28Models ranked992

28 leaderboards · 992 models ranked

LMArena WebDev Arena what it measures, how it is scored and who leads it

Arenas · LMArena · Arena rating

Leaderboard

Scores from LMArena WebDev Arena, published Nov 10, 2026

RankModelArena rating
1Claude Fable 5.2Anthropic · as “Claude Fable 5.2 (xHigh)”1,836
2Claude Opus 5.5Anthropic · as “Claude Opus 5.5 (Max)”1,814
3Claude Sonnet 5.5Anthropic · as “Claude Sonnet 5.5 (xHigh)”1,773
4Gemini 4 CarbonGoogle · as “Gemini 4 Carbon (High)”1,768
5GPT-6 AstraOpenAI · as “GPT-6 Astra (Max)”1,764
6GPT-6.1 SolOpenAI · as “GPT-6.1 Sol (Max)”1,743
7Claude Fable 5.1Anthropic · as “Claude Fable 5.1 (Max)”1,742
8Muse Wonder FullMeta · as “Muse Wonder Full (Max)”1,723
9Claude Opus 5Anthropic · as “Claude Opus 5 (Max)”1,691
10Kimi K3.1 ProMoonshot AI · as “Kimi K3.1 Pro (Max)”1,688
11Gemini 4 ArgonGoogle · as “Gemini 4 Argon (High)”1,682
12GPT-6.1 TerraOpenAI · as “GPT-6.1 Terra (Max)”1,679
13Qwen3.8 Max (0902)Alibaba · as “Qwen3.8-Max (0902)”1,676
14GPT-6 SolOpenAI · as “GPT-6 Sol (Max)”1,673
15DeepSeek V4.1 FlashDeepSeek · as “DeepSeek-V4.1-Flash-20261017 (Max)”1,664
16Kimi K3Moonshot AI · as “Kimi K3 (Max)”1,658
17Muse Spark 1.3Meta · as “Muse Spark 1.3 (Max)”1,655
18Muse WonderMeta · as “Muse Wonder (Max)”1,649
19Inkling 2 SmallThinking Machines · as “Inkling 2 Small”1,646
20Grok 4.7SpaceXAI · as “Grok 4.7 (xHigh)”1,641
21MiMo-V2.6-FlashXiaomi · as “MiMo-V2.6-Flash”1,638
22Hy4 previewTencent · as “Hy4 preview”1,632
23Claude Fable 5Anthropic · as “Claude Fable 5 (High)”1,625
24MiMo-V2.6-ProXiaomi · as “MiMo-V2.6-Pro”1,624
25Qwen3.8-Flash-NextAlibaba · as “Qwen3.8-Flash-Next”1,624

About this benchmark

Task
Build a web app from a prompt; people vote for the better of two.
Dataset
Crowd-sourced prompts and votes.
Method
Pairwise votes fitted with a Bradley-Terry model.
Metric
Arena rating
Organization
LMArena
Official leaderboard
arena.ai/leaderboard/code

Sources: each benchmark's paper (arXiv) and its official site or leaderboard. Scores are the boards captured here every day, exactly as published.